Knowledge Representation & Reasoning

3. KR&R within perception-action loops

In order to extend the knowledge representation and reasoning techniques we have discussed so far to reason about how to execute actions we will now consider more realistic representations of perception-action loops. This approach takes the view that a computational model for controlling a robot agent accomplishing complex manipulation tasks has to explain how vague instructions such as "pour the popcorn onto the plate'' can be translated into the body movements that the robot has to generate. To infer such a body motion the robot the robot needs to infer that it has to grasp the pot by the handles, hold the pot horizontally, tilt the pot around the axis between the handles, remove the lid before pouring, and so on.